FMURF Project Overview:

Project Overview: 

We rely on Automatic Speech Recognition (ASR) technologies such as Google Assistant and Apple’s Siri every day. While their usage and accuracy have been examined in standardized, monolingual speech and in English, the participation of bilingual groups is understudied, and if they are included, only one of their two languages is considered. For this project, we aimed to understand how well ASR systems process the speech of Spanish-English bilinguals. To explore this, we recruited 16 Spanish-English bilinguals whose first language was either Spanish or English with an average age of 23 and who learned their second language at around the age of 8. We played recordings of 60 commands and a short paragraph from each participant in both Spanish and English to two ASR systems: Google Assistant and Siri. Across the two languages, preliminary results showed that Siri made less errors while following commands and while transcribing speech from bilinguals whose first language is Spanish in contrast with those whose first language is English. Furthermore, when listening to English, the difference in errors made by Siri among these two groups of bilinguals was smaller. By focusing on bilinguals—a community often overlooked in tech development—this research aims to pave the way for more inclusive technologies that serve linguistically diverse populations.